Full Text of - Virtualization at Fujitsu

5/9/2019 Full text of "Virtualization at Fujitsu"
Full text of "Virtualization at Fujitsu"

See other formats
https://archive.org/stream/WpVirtualization/wp-virtualization_djvu.txt 1/81
White paper Virtualization at Fujitsu
FUJITSU
White paper
Virtualization at Fujitsu
This whitepaper examines the various virtualization technologies and shows the range of ICT
infrastructure layers where they are applied. It also shows how Fujitsu incorporates these
technologies into end-to-end solutions as an essential part of its Dynamic Infrastructures strategy.
Page 1 of 32
www.fujitsu.com
Introduction
What is virtualization?
Broadly speaking, virtualization in an information and communication

technology (ICT) environment is about breaking the dependencies
between the various layers in the infrastructure. Introducing a
virtualization layer enables abstraction from the complexity of the
underlying layers. However while it hides the complexity it also
mediates access to it, allowing the creation of virtual resources that can
be mapped to the underlying physical resources. This definition is
applicable at any level where virtualization is applied.
Virtualization is not new
Virtualization, introduced more than 30 years ago with mainframe

systems, initially pursued two basic objectives:
■ With server hardware extremely expensive at the time, it needed to

be fully utilized to gain maximum cost benefit. Server virtualization
enabled grouping of the various operating system (OS) instances on
the same physical hardware system. That significantly improved
flexibility and capacity utilization.
■ By separation of the application reguirements from the restrictions of

the physical hardware and software infrastructure layers,
virtualization enabled further development of new technologies,
without impact on the existing software infrastructure.
The technology developed at that time corresponds very closely to the

current approach to server virtualization using modern day hypervisor
technology.
Server virtualization is rooted in mature mainframe technology
Fujitsu was one of the virtualization pioneers, and has decades of

experience in providing and managing virtual machine infrastructures.
Fujitsu GlobalServer and BS2000 mainframes are perfect examples of
continuity and long-term investment protection in mission-critical
datacenter environments brought about by virtualization. This ongoing
development has allowed resource allocation to occur dynamically at a
very granular level. It has also meant considerable reductions in energy
consumption, space and administrative effort. In addition, it has
achieved high availability in a cost effective manner, with services
made ready much faster. A virtualized approach also enables parallel
operation of production, test, and development systems allowing both
pre- and post-production systems to exist on the same hardware.
Monitor
Application VM
= OS+App
Customer
Customer
Application VM
= GS+App
Today the approach to virtualization and the levels at which it is

applied are more diverse. The following diagram shows the various
infrastructure levels of virtualization deployment and the solutions that
are freguently used.
Application
abstractions
Desktop
abstractions
Container
abstractions
OS-basedVM
abstractions
Hypervisor-based
VM abstractions
Physical
abstractions
Server I/O
abstractions
Network
abstractions
Storage
abstractions
1 Microsoft SoftGrid, Altiris Software Visualization Solution,

1 Ardence, SAP, Oracle...
Hmj^HmH VMware View, Citrix XenDesktopServer, ...

VwdHpiH Solaris Zones, SWSoft.Virtuozzo/OpenVZ, Java EE...
1 VMware {GSX) Server, Microsoft Virtual Server, KVM...
1 VMware ESX Server, Xen, Microsoft Hyper-V
1 Intel VirtualizationTechnology(VT),
1 AMDVirtualization(AMD-V)...
BpH 1
HM | tisco, Brocade...
1 Fujitsu ETERNUSCS, Fujitsu ETERNUSDX, EMC, NetApp,

1 Veritas, Brocade, Microsoft,..
The wide range of virtualization solutions however means that a key

task of ICT managers is identification of the right technology for each
specific reguirement, and integration of that into the ICT infrastructure.
However, the diagram clearly shows that no single virtualization layer
will likely meet all reguirements in a state-of-the-art dynamic
infrastructure. Instead, a combination of the various technologies is
reguired to create an ideal customer-specific solution.
This white paper discusses the best-known technologies, shows their

advantages and describes their connections with other layers and
solutions. It presents the technologies that Fujitsu uses, as well as
those it has developed in-house, to offer each customer a consistent,
end-to-end, virtualization solution.
Page 2 of 32
www.fujitsu.com
Server virtualization with x86 servers
No other server architecture has gained in importance over the past

decade as much as x86-based server platforms. The complete Fujitsu
PRIMERGY product family is the embodiment of this form of
architecture, and based today on virtualization support in processor
technology. The rapid pace of this technology results in an
unprecedented 10" price/performance ratio. In fact, other components
have trouble keeping up with the speed at which the hardware is
developing. That results in special reguirements for this platform,
which are tackled by virtualization.
The traditional x86 platform approach
The architecture of a traditional industry standard server consists of

three important layers.
■ First is the hardware layer, which is a set of different devices and

compute resources.
■ Second is the OS layer, which has overall control of hardware

operation. It is responsible for managing the hardware resources and
sharing them among the different tasks, based on priorities.
■ Third is the applications/services layer, which uses the OS as an

interface to access the reguired hardware resources.
In this traditional platform approach, each configuration layer is tightly

coupled to each preceding (or lower) layer. As a result, hardware and
software lifecycles are strongly dependent on each other. This means
for example, a legacy application that runs on an OS not supported by
new hardware cannot upgrade to that new hardware. As hardware
changes regularly, this becomes a serious issue.
Due to historical reasons, OS resource management is not very efficient

and applications can interfere with each other over joint use of
particular resources. This can lead to servers, sized and installed for
peak loads, running only a single application. This resulting
over-provisioning of system resources leads to correspondingly poor
sever utilization. Poor server utilization, in its turn, leads on to an

explosion in server numbers, high energy consumption, additional
heat generation, expansive floor-space reguirements, and of course
higher expense on maintenance and support. Likewise, scalability is
limited dynamically and high availability becomes difficult and
expensive to implement,
The overall result is management complexity and low efficiency, with

considerable effort needed to deploy new services; i.e. the time
reguired to respond to changes in business is too long.
x86 platform virtualization
To overcome the traditional challenges created by the strong

dependencies between hardware and OS, a decoupling must occur. The
use of hypervisor or "server virtualization" technology achieves this.
By definition, server virtualization means hiding the server resources,

including the number and the identity of each physical server,
processor, and OS, from the end users. In particular, applications and
services now operate independent of the dedicated hardware
resources. This enables more flexible control of both hardware and
software resources, and improves the value proposition across a wide
range of ICT requirements. According to a range of studies,
virtualization will be a pervasive and important trend in servers
through 2014.
Application
® ® ° 5 §> ®
UiardwanP
Traditional approach
Application
OS
Virtualization layer
Hardware
Virtualization approach
Hypervisor technology lets multiple instances of an OS and associated

applications run on the same physical hardware system. It also enables
the consolidation of large numbers of under-utilized physical servers
onto fewer well-utilized servers. Virtualization creates independence
between the software and hardware layers and becomes the
foundation for new operating processes.
In the past, the ability to run older OS versions on the latest hardware
was one key motivation for deploying hypervisor technology. Now
customers also expect further benefits from server virtualization; in
particular greater agility, higher availability and lower operating costs.
The traditional platform approach therefore becomes synonymous with

a high Total Cost of Ownership (TCO).
Page 3 of 32
www.fujitsu.com
Virtual machines in OS hosted environments
Some virtualization technology is used less with today's x86 servers. OS

hosted environments are an example. Here the virtualization layer sits
on top of the host OS, and is just one of the many features of the
physical server's OS. In this case, the virtualization layer has the status
as any other application running on the hardware and OS. It also
exclusively uses the device drivers and resource management of the
host OS. Examples are VMware Server and Microsoft Virtual Server.
The virtual machine concept in OS hosted environments came into

being at the early stage of server virtualization. One of the main areas
of application was the ability to run different development and test
systems on the same hardware. Today this is only relevant in small
system environments.
App
App
OS
App
OS
App
OS
Virtualization Layer (VMM]
Hypervisor
Host OS (Windows or Linux)
Hardware
Mid-sized businesses have the same expectations, but their top five
implementation priorities are different:
■ Consolidation: This is number one here too, but the level of savings is
lower due to the smaller server deployment scale.
■ Protection: Number five with large enterprises, it rises to number two

with mid-sized businesses. Here virtualization and low-cost Storage
Area Network (SAN) are often combined. This can significantly reduce
the cost and complexity of implementing disaster recovery.
■ Deployment: Most mid-sized businesses have limited administrative

resources. Virtualization helps reduce administrative effort and
increases speed of deployment.
■ Freedom of choice: Virtualization allows purchases based on

competitive pricing without worrying about the overheads of
supporting multiple vendors.
■ Operational agility: Here virtualization helps in adapting server

resources to changes in workload demands. While not at the same
level as with large enterprises, implementing new services at branch
offices, for example, becomes much easier.
Overall, mid-sized businesses, while also driven by the promise of

consolidation and cost savings, see virtualization as an opportunity to
introduce solutions that would otherwise be difficult and expensive to
deploy.
Bare-metal environment
OS hosted environment
Customer expectations of the virtual machine concept
When it comes to the question of what customers expect from the

virtual machine concept, we have to distinguish between large
enterprises and mid-sized companies. While the overall perspective of
both can appear the same, their priorities are often different. With
large enterprises the top five priorities are as follows:
■ Consolidation: Which provides the opportunity for dramatic capital

equipment cost savings.
■ Virtualization: As an enabler in reducing the administrative burden of

loading and configuring the endless numbers of new servers arriving
on the loading dock.
■ Operational agility: The ability to move workloads from one server to

another as changes demand, with reduced administrative effort.
■ Freedom of choice: Virtualization makes it easier to accommodate a

mix of vendors and server models, within the datacenter.
■ Protection: Virtualization simplifies disaster recovery - but only with

x86 platforms.
However due to the heterogeneous nature of most large enterprise ICT

infrastructures, replication for disaster protection tends to be complex.
That is why protection has the lowest priority in their top five. In
general, it is fair to say, large enterprises tend to focus on the cost
savings associated with virtualization.
Page 4 of 32
www.fujitsu.com
Virtual machines
By definition, a virtual machine is a software-driven implementation of

a computer. One that executes programs in an identical way to a
physical machine but which runs as an isolated environment as part of
the physical computer. A virtual machine is also described as a guest,
with the physical computer as the host. As numbers of virtual machines
can run on a single physical machine, the physical machine will appear
to users to behave like many servers.
Server virtualization tricks the virtual machine into only seeing the
same virtual hardware (CPU, memory and I/O devices) regardless of the
overall underlying equipment in the physical machine.
Key features
Four key features characterize the virtual machine concept.

First, virtual machines are fully isolated from each other. Isolation is
essential to ensure process availability and maintain security. If for
instance a virtual machine fails ("blue screen"), there must be no
impact on other virtual machines running on the same physical
hardware.
Second, the overall physical server hardware is partitioned into several

virtual machines, each running its own (guest) OS. Each guest OS may
be quite different. Hardware resources, such as CPU, main memory or
I/O devices, are shared among the virtual machines. A technique called
encapsulation saves the status of each virtual machine as a file. Simply
by copying this file, the virtual machine is transferred quickly and easily
from one physical environment to another.
Third, with the hypervisor situated between the software stack

(containing the OS and the application) and the server hardware,
software and hardware can operate completely independently. This
enables running of legacy applications on the latest hardware
technology, even though the OS needed to run the application does
not support the necessary drivers on the new physical hardware. The
hypervisor acts as an intermediary and ensures the investment
protection of legacy applications.
Case studies
Major university in Tokyo
With three separate campuses in the greater Tokyo area, this science
and technology university was seeking a way to reduce their total
cost of ownership of ICT equipment. The existing UNIX servers at
each location required separate administration as well as local
Network Attached Storage (NAS). Introduction of 500 Mbps
inter-campus transmission speeds paved the way for a more
consolidated approach. The move to on-demand remote class work
has also put pressure on the inflexibility of the existing systems.
Initially they envisaged a virtual machine environment at each
campus. However, with the new communications backbone this
quickly developed into a centralization plan.
Extensive testing and evaluation lead to the adoption of a Linux

environment based on Fujitsu PRIMERGY rack servers and VMware
plus data consolidation in a Storage Area Network (SAN) based on

Fujitsu ETERNUS DX storage, with VMware vMotion for data
migration. The result was of the migration of 20 UNIX servers onto
just 1 2 Red Hat Linux servers. In addition, the university discovered
that the new centralized environment had the capacity to handle 80
plus virtual machines. This would also support their future growth,
with fast deployment and no disruption to existing services. The
success of this work and the enhanced fault tolerance of the new
infrastructure have drawn great attention from other major private
universities.
Server virtualization at Finish investment bank
This Finnish investment bank runs its own in-house developed

applications at brokerage offices around the world. However, in the
past, based on their perception of the best local support, each of the
local offices procured the required hardware independently. The
result was a rather heterogeneous server landscape. This caused
major problems and high administration costs, particularly when
new applications needed delivery for testing and local
implementation.
Fourth, all virtual machines are contained in a common file system,

e.g. VMFS (Virtual Machine File System). This permits the building of
failover clusters between virtual machines. This is much more
affordable than clustering using physical servers; which are
traditionally expensive, complex and difficult to manage.
Now using virtualization technology, the Bank performs quality

assurance and testing at their headquarters. This means a significant
saving in development and system assurance costs. Deployment also
occurs easily and rapidly, as headquarters staff simply copies the two
binary files of the tested virtual machines, and remotely deliver them
to the target systems, wherever they are worldwide. This is possible
because virtualization separates the requirements of the applications
from the underlying hardware resources. New services are
deployable within minutes and if a disaster occurs, fast recovery
using the system backup copies is quick and simple. Now everything
works with minimum administration effort.
Page 5 of 32
www.fujitsu.com
Hypervisors
General characteristics
There was for a long-time detailed debate on the best type of

hypervisor architecture. Which created the lowest overhead, and what
virtualization method was the most efficient?
"Para-virtualization", "full virtualization", "trap and emulate" and "binary

translation" (see page 21 ) are some of the buzz words used in this
connection. However once standard processor technology started to
support virtualization, the debate was pushed to the background. Most
hypervisors now use these "hardware assist" processor interfaces and
all achieve comparable performance. Para-virtualization, where some
time-critical functions are performed on the host rather than in the
virtual machine, is now only used in exceptional cases; in particular,
where the hypervisor vendor also has unrestricted access to the OS.
In addition to this standardization of virtualization methods, all

vendors offer functions that result in higher availability and agility, as
well as efficient management of the virtual environment. All hypervisor
vendors offer a special management suite for virtualization. Each
offering differs in terms of its management coverage and end-to-end
solutions and this has become the prime criteria for choosing a
suitable hypervisor. As a result, the basic hypervisor has become a
commodity component. So, in comparing the price of different
products, it is essential to analyze the product structure and the range
of associated functions closely, to obtain truly comparable results.
Market relevance
Today over 20 hypervisors are available in the market, but the direction
is towards a few dominant hypervisor vendors. It is for this reason that
Fujitsu primarily focuses on a select few hypervisors for its
consolidation solutions. The following graphic from the Gartner
research company, shows the projected market share for virtualization
vendors in 2012, by number of VMs deployed.
Red Hat
2% A Others 1%
Citrix6%_ \ |
Microsoft
27%
Vmware
64%
FY 201 2 (58MVMs)
Note that major ISVs now offer their products as appliances, i.e. the
application ships as part of the virtual machine and not in conventional
installation format. Conseguently, the application is independent of
the type of server it runs on. This reduces the time and effort ISVs need
to spend on Q&A. However, it can result in end customers using more
than one hypervisor. To minimize the complexity this involves, vendors
are working on standardizing VM formats (for example Microsoft and
Citrix) enabling virtual machines of a specific hypervisor to run on
another hypervisor, at no cost or effort. This feature is also important
when choosing a suitable hypervisor.
Note: The seguence of product explanations that follow in the next

section is not in exact market share seguence. Instead, it follows the
chronological sequence of product development and dependency.
Virtualization technology in the x86 market started around the year

2000, with client and server offerings from VMware. The initial
technology was based on software emulation (binary translation).
Then in 2004/2005, processor vendors AMD and Intel implemented
basic software aware hypervisor functions in the hardware layer. This
meant CPU and memory functions, previously handled in the software
stack, now became available to hypervisor vendors as a "function call".
This was the catalyst for other hypervisor vendors to enter the market
with new offerings based on the "hardware assist" technology of the
processor vendors.
Page 6 of 32
www.fujitsu.com
1 . VMware vSphere
VMware has been an influential force in shaping server virtualization

over the past decade. VMware's virtualization technology began with
the PC (GSX). For the first time, multiple OS could run on the same PC
allowing more flexible test environments. It was not long before this
method was also used with servers.
As a result, the focus shifted to server consolidation and the possibility

of running aging Windows NT4 environments on the latest hardware.
With its vCenter management software, VMware set new standards in
the management of virtualized environments. High availability (HA),
load balancing within vSphere server farms, distributed resource
scheduling (DRS), and above all vMotion, which lets virtual machines
move during operation from one physical instance of the vSphere farm
to another, were all pioneering developments. They also had a lasting
influence on the planning of other vendors. The next step was fault
tolerance (FT) which enables continuous operation by mirroring virtual
machines on another vSphere host. Interaction between layers has also
improved with many built-in Application Programming Interfaces
(APIs). Security software vendors, Storage vendors and Network
vendors (Cisco Nexus 1000), can now use these APIs (VMsafe, vShield)
to deliver more highly optimized infrastructures.
Overview of the vSphere hypervisor
The current "classic" VMware architecture looks like this:
Service console
vCenter SDK 3 d party

agent agent agents
VM
VMM
VM VM VM
VMM _ VMM _ VMM
The VMware vSphere hypervisor runs directly on the hardware layer and
acts as an independent OS. The kernel takes care of all of the following:
■ Management of the VMs
■ Separation of the VMs from each other
■ Efficient distribution of resources between VMs in line with

administrator requests
■ Mapping of machine instructions issued in a VM
■ Mapping of I/O jobs
The overall size (2GB) of the classic main version is due to the Linux
subsystem. The subsystem enables the installation of agents to
manage the underlying hardware. It also includes interfaces for
management of the hypervisor. VMware intends to phase out this
architecture in 201 1 and reduce the hypervisor back to its pure kernel
stack, with special interfaces for management. The future embedded
VMware architecture looks like this:
CIM
Client
VM VM VM VM
v5phere API
vSphere Client
vCLl, Power CLI
_ iostc
1 CIM 1
Provider 1
Network
stack
Virtual 1 n 1 Clustered 1 r
NICsS 1 °™ ce 1 VM file 1 5tora f 1
c t u 1 drivers 1 1 stack I
Switches 1 1 system 1
m '-,-oe H
vCenter
VMware developed this kernel, with its impressive lean design

requiring just 70MB, in-house. The results will be:
■ No further need to support the Linux subsystem
■ Greater security and protection from external attacks
■ Easier implementation on OEM hardware with cost and efficiency

benefits for users.
Info box
Common Information Model (CIM)
With the removal of the subsystem in which agents can run, a new
interface is required. This now becomes the task of each hardware
vendor using the new Common Information Model (CIM) Interface.
CIM is an open standard that defines how computing resources are
represented and managed. It enables a new framework for
agent-less, standards-based, monitoring of hardware resources for
the vSphere hypervisor. This framework consists of a CIM object
manager, otherwise called a CIM broker, and a set of CIM providers.
Each hardware vendor will need to create specific CIM providers for
their hardware.
CIM providers are mechanisms that enable management access to

device drivers and underlying hardware. In fact hardware vendors
(both server manufacturers and specific hardware device vendors),
will now have the flexibility to write CIM providers that best monitor
and manage their particular devices. VMware is also writing
providers that implement monitoring of server hardware, storage
infrastructure, and virtualization-specific resources. These providers
run inside the vSphere system and are designed to be extremely
lightweight and focused on specific management tasks. The CIM
object manager in vSphere hypervisor has a standard interface.
Developers use this to plug-in new providers. However, the providers
must be packaged with the system image, and cannot be installed at
run time. The CIM broker takes information from each CIM provider
and presents it to the outside world via standard APIs, including
WS-Management
Page 7 of 32
www.fujitsu.com
The lean kernel design, in particular, distinguishes VMware vSphere

from other hypervisor architectures. VMware no longer needs the
support of another operating system instance to handle its tasks. This
makes it independent of any restrictions that may occur with such
instances.
However, interaction with other OS instances (see section on Citrix

XenServer and Microsoft Hyper-V) can have advantages, in particular
with I/O handling. VMware vSphere hypervisor is the connecting link
between the VMs and the underlying hardware. Each VM only "sees"
the formally supported hardware (SCSI disks, Network Interface Cards
(NICs), processors and storage; however, this does not always
correspond to all existing hardware the hypervisor has to serve. To
address all hardware from the hypervisor, drivers exist for most
common hardware components. However, caution is reguired if special
hardware components are needed for an application (e.g. security
dongles, ISDN cards, special graphics support, etc.). It is then necessary
to check on a case-by-case basis whether VMware supports the relevant
drivers. Manufacturers of each special hardware component must
ensure they certify with VMware for the servers in guestion.
Manufacturers not on the VMware Hardware Compatibility List (HCL)
cannot expect support from VMware in the event of an error.
VMware management
As mentioned at the outset, management of virtualized infrastructures

is a core component in every hypervisor scenario. The following shows
the core functions of VMware vCenter:
VMware vCenter focuses on managing the virtual machines (VMs). In

line with VMware's philosophy, real server instances no longer play a
major role. The real instance, the VMware server, is viewed as part of
the server farm. The resources, the server contributes to the farm, are
comprehensively administered by the relevant management layers,

and divided among the VMs as specified. Individual servers no longer
play a role in this distribution. VMware vCenter allows a new server to
be included in a server farm with very little effort. It also enables
simple replacement of servers that have failed due to a hardware fault.
However, VMware vCenter does not enable the provisioning of
substitute physical servers. This reguires the addition of automated
processes or products, such as Fujitsu ServerView Resource Coordinator
VE (See separate section in this white paper.)
VMware provides a comprehensive range of proven technologies that

enable end-to-end solutions. VMware has a wide ranging and powerful
stack of solutions including, site recovery manager, fault tolerance,
data recovery, thin provisioning, distributed virtual switching and
dynamic power management. Certainly, VMware has the broadest
portfolio of solutions of any hypervisor.
Vmware vCloud
■ vCloud Director
■ vCloud Request Manager
■ vCloud Express
VmwarevFabric
— vFablit tc Server
vFablk Hyperic "
vFabric GemFire
vFabric Enterprise Ready Server o
vFablric RabbittMQ -£
i Centralized management interface
i VM provisioning with templates
i Performance monitoring
i Secure access control
i SDK (Software Development Kit) for automation
Vmware vCenter server
vCenter Qrchestrator
vCenter Server Heartbeat
vCenter Capacity IQ
vCenter AppSpeed
vCenter ChargeBack
vCenter Site Recovery Manager
vCenter Configuration Manager
Vmware End User Computing
■ VMware View
■ VMware ThinApp
■ VMware Workstation
■ VMware Fusion
■ VMware ACE
■ Zimbra
■ VMware Mobile Virilization
Vmware vSphere
Virtualize and aggregate
hardware resources
Built-in service level controls for
applications
VMware vShield
— vShield Endpoint
vShield App
vShield Edge
Page 8 of 32
www.fujitsu.com
VM migration technology
One great advantage of server virtualization is the ease of handling

virtual machines. Virtual machines are handled as files in a common
cluster file system. This independence from the underlying server
hardware makes it easy to (re)start a virtual machine on every
virtualized host system running the same hypervisor.
Agility is the next benefit, in particular this covers the ability to move a
virtual machine from one server to another while it is running. All
hypervisor vendors deliver this functionality under various names
including, vMotion (VMware), XenMotion (Citrix) and Live Migration
(Microsoft). Here we will use a vMotion example, but the technology
and functions of each are very similar.
VMware vMotion
Three underlying technologies enable the live migration of a virtual

machine from one physical server to another.
First, the entire virtual machine is encapsulated in a set of files on

shared storage (Fibre Channel or iSCSI connected SAN or Network
Attached Storage (NAS)). Then VMware vStorage VMFS allows multiple
installations of VMware vSphere to access the same virtual machine
files concurrently.
Secondly, the active memory and precise execution state of the virtual
machine is rapidly transferred over a high-speed network. This lets the
running virtual machine switch almost instantaneously from the source
vSphere host to the destination vSphere host. Users do not notice.
vMotion enables fast switchover by keeping track of on-going memory

transactions using a bitmap. Once it has copied the entire memory and
system state to the target vSphere host, vMotion suspends the source
virtual machine, copies the bitmap to the target vSphere host, and
resumes the virtual machine on the target vSphere host. Using a
separate Gigabit Ethernet a vMotion process takes a few minutes.
User benefits
vMotion enables live migration of virtual machines between physical

servers with zero downtime. This means service continuity and
complete transaction integrity. vMotion becomes a key enabling
technology for creating dynamic, automated, and self-optimizing
datacenter. Administrators can also move planned server downtime
from weekends back into normal business hours without impact on
applications or users. This helps reduce administration and
maintenance costs. Today over 40% of hypervisor users benefit from
this live migration technology. User benefits can be summarized as:
■ Optimized workloads
■ Ability to perform maintenance during normal working hours,

without costly out-of-hours overtime
■ No negative impact on performance
App
1 |
■ I
App
OS
._.
App
OS
VMotion technology
vSphere hypervisor
Hardware
vSphere hypervisor
Hardware
Thirdly, the underlying vSphere host virtualizes the networks use by the
virtual machine. This ensures that even after migration, the network
identity and network connections of the virtual machine are preserved.
vMotion manages the virtual Media Access Control (MAC) address as
part of this process. Once it has activated the destination machine,
vMotion simply pings the network router to ensure it is aware of the
new physical location of the virtual MAC address. Since the migration of
the virtual machine using vMotion preserves the precise execution
state, network identity, and active network connections, the result is
zero downtime and no user disruption.
Page 9 of 32
www.fujitsu.com
2. CitrixXenServer
Xen was the Linux community's contribution to virtualization, But first,

it is worth looking at how this technology is offered to the marketplace.
Xen was the Linux community's contribution to virtualization.
Developed at Cambridge university the base of the product is available

under the familiar rules of the open source community. XenSource, a
company founded by developers from that university, then took charge
of further hypervisor development. XenSource also provide a charged
for Xen distribution version of its own with far more functionally than
the original. All major Linux vendors (and more recently ISVs like
Oracle) have also developed their own derivatives based on the open
source Xen version. Each licenses their product using a different model.
Citrix acquired XenSource in 2008 and now develops Xen as part of its
server and desktop virtualization strategy. This acquisition also resulted
in a dampening effect on the uptake of the free versions. The figure
below shows the basic mode of operation of a Xen hypervisor
As with VMware, the actual Xen hypervisor runs directly on the

hardware. Unlike VMware however, the hypervisor does not handle the
"important" functions. These are handled by what is termed "domain 0."
This privileged virtual machine runs the standard OS of the particular
Linux vendor. The OS then controls each virtual machine (called in this
case domain 1, 2, to domain n). Domain is especially important for
I/O handling.
Every I/O issued, for virtual hardware from a virtual machine, is

internally redirected to domain 0. I/O is then processed on the real
hardware using the original drivers included in each particular Linux
distribution. This lowers the costs for OEM's certifying such hypervisors.
Xen supports paravirtualization and full virtualization.
XenServer features
XenServer features can be divided into four function groups:
DomO
Linux
Doml
Linux
Dom2
■\e:wa~e
DomO - Management domain for Xen guests
■ Controls compute resources dynamically (e.g. memory, CPU, I/O)
■ Provides interfaces to the physical server
■ Provides administration tools and interfaces
Doml, 2, etc.- Guest OS
■ Hosts the application workloads
■ Typically uses virtual device drivers to connect to the physical drivers

in DomO by the hypervisor
■ Can also use physical device drivers directly
■ Can be stored in a file image
Hypervisor
■ Is the virtual interface to the hardware - virtualizes the hardware
■ Manages conflicts caused by OS access to privileged machine

instructions
■ High performance virtual infrastructure
■ Datacenter automation
■ Advanced integration and management
■ Citrix VMLogix
High performance virtual infrastructure
The high performance virtual infrastructure set is free and builds the
main base for virtualization, including the following function set:
■ XenServer hypervisor
■ XenMotion
■ XenCenter multi-server management
■ Conversion tools
The promotion of XenServer, based on the open-source design of Xen,

is as a highly reliable, available, and secure virtualization platform. The
use of 64-bit architecture allows it to provide near native application
performance and unmatched virtual machine density. The intuitive
wizard-driven utility ensures easy server, storage, and network setup
and is referred to as "Ten Minutes to Xen" in the literature. Recovery is
simplified as disk snapshot and revert create periodic snapshots of both
virtual machines and data. This allows recovery to a known working
state after a failure. The ability to clone disk snapshots also speeds
system provisioning.
XenMotion enables the movement of workloads without user

interruption while the workloads remain running.
Page 10 of 32
www.fujitsu.com
Citrix XenCenter provides all the VM monitoring, management, and

general administration functions in a single interface including
configuration, patch management, and virtual machine libraries. This
lets ICT staff manage hundreds of virtual machines from a centralized,
management console installed on any Windows desktop. Resilient
Distributed Management Architecture distributes server management
data across the servers in a resource pool to ensure that there is no
single point of management failure. If a management server fails, any
other server in the pool can take over the management role.
Conversion tools enable conversion of physical and virtual servers (P2V

and/or V2V) for faster setup and migration, and makes transformation
to XenServer nearly instantaneous. Included in XenServer, they can
convert any physical server, desktop workload, or existing virtual

machine, to a XenServer virtual machine.
Datacenter automation
Datacenter automation includes the following function set:
■ Site recovery
■ High availability
■ Auto-restart capability
■ Dynamic workload balancing
■ Host power management
■ Lifecycle management
■ Memory optimization
Site recovery provides site-to-site disaster recovery planning and

services for virtual environments. This is easy to set up and fast to
recover, and provides the important ability to test disaster recovery
plans to ensure they remain valid.
High availability automates the restart of virtual machines following a

VM, hypervisor, or server level failure.
Auto-restart capability protects all virtualized applications and enables

higher levels of business availability.
Dynamic workload balancing provides automatic balancing of virtual

machines within a resource pool. This improves system utilization and
increases application performance. By matching application
reguirements to available hardware resources, this function also places
VMs on the most suitable host in the resource pool.
Host power management uses embedded hardware features to lower

datacenter electricity consumption. It dynamically consolidates VMs on
fewer systems and then powers off unutilized servers as demand for
services reduces.
Lifecycle management provides administrator access to defined virtual

environments using a predefined set of workflows. This helps reduce
complexity as well as the time and cost to develop, maintain, and
deliver applications.
Memory optimization shares unused server memory among the VMs

on the host server. This increases workload density by enabling more
virtual machines per server. It also has a cost reducing effect as well as
improving application performance and protection
Advanced integration and management
Advanced integration and management includes the following

function set:
■ Provisioning services
■ Role-based administration
■ StorageLink
■ Performance reporting and alerting
■ Heterogeneous pools
■ Live VM snapshot and revert
Provisioning services reduce storage reguirements by creating a set of

"golden" images. These can then be streamed to both physical and
virtual servers for faster, more consistent, and reliable application
deployment.
Role-based administration helps improve security, and enables

delegated access, control, and use of XenServer pools. It also maintains
a tiered access structure with different permission levels.
StorageLink provides integration with leading network storage

platforms. This makes it easier for storage managers to leverage their
existing data management tools to maintain consistent management
processes in both physical and virtual environments. Key data
management processes can be created from a single StorageLink
interface.
Performance reporting and alerting enables instant notification with

historical reporting of VM performance. This enables rapid
identification and diagnosis of faults in the virtual infrastructure.
Heterogeneous pools means resource pools can contain servers with

different processor types, and support XenMotion, High Availability,
workload balancing, and shared storage functionality.
Live VM snapshot and revert, captures memory and disk state during a
snapshot. This speeds system provisioning and ensures periodic
backup.
Page 11 of 32
www.fujitsu.com
CitrixVMLogix group
The Citrix VMLogix group includes the following function set:
■ Web self-service for Citrix XenServer
■ VMLogix and Citrix OpenCloud
VMLogix integrates Web Self Service, a simple web-based console, into

XenServer. Web Self Service lets ICT administrators delegate individual
virtual machine (VM) administration rights to application owners and
enables application owners to manage the day-to-day operations of
their VM's. This makes it easier for ICT teams to build, share and deploy
production-like environments on-demand in both private and public
clouds. It also allows migration of virtual workloads between
production stages. By providing self-service access to a single pool of
computing resources, cloud providers can help users reduce their
capital expenditure and improve flexibility, even across diverse
virtualization and cloud environments.
Citrix XenServer with Citrix VMLogix adds key lifecycle management

capabilities to the Citrix OpenCloud platform. This makes it easy for
cloud providers to offer infrastructure services that extend from
pre-production, to staging, and then deployment, ensuring business
continuity and guality assurance. VMLogix provides support for
multiple virtualization platforms. This removes the barriers between
on-premise and cloud, making mobility of workloads as easy as
operation within a local pool.
XenServer Desktop Integration
XenSource was purchased by Citrix at the beginning of 2008. As part of

this, Citrix also took over the management product XenCenter. The
functionality of this management product had set the standard in the
Xen arena. Now Citrix has developed a new strategy around the Xen
product positioning XenServer as the elementary component of this
strategy.
The focus is Citrix's solution for desktop virtualization. Deep integration

with XenApp Server has resulted in Citrix expanding its server-based
computing portfolio to include dedicated virtualized desktops. This
integration leverages, without restriction, the advantages of the
well-tested ICA (Independent Computing Architecture) protocol
together with all access paths and the management tools in the
conventional Citrix infrastructure. At the same time, Citrix is positioning
its virtualization product XenServer as the solution for virtualization of
traditional Citrix Presentation Servers and Microsoft Terminal Servers.
This will offer a simpler means for future scaling of these servers. Due
to limitations in the Windows kernel, a single Presentation/Terminal
Server instance was only able to use the capacity of new servers to a
limited extent. However, with the aid of virtualization, multiple
instances of such a server can now run on the same hardware. This
result means a sharp increase in the number of concurrent users able
to be supported. This optimization was possible only with close
cooperation with Microsoft, another major vendor of virtualization
technology. Citrix, as a result, is expanding its management portfolio
for the Microsoft hypervisor.
Page 12 of 32
www.fujitsu.com
3. Microsoft Hyper-V
Microsoft released its hypervisor product in mid-2008 and provided a

new version, Microsoft Hyper-V 2.0, in October 2009. Prior to 2008,
Microsoft had offered Microsoft Virtual Server, a free-hosted virtual
machine product that ran on top of Windows Server OS. Looking at the
architecture of Hyper-V, there is some similarity to the previously
explained Xen architecture:
Microsoft Hyper-V Server 2008 R2 is available as a free downloadable
product. It enables execution of many virtual machines on a single

server. In addition, key functions - like live migration and high
availability - are available. Hyper-V Server 2008 R2, which runs in the
parent partition, consists of the hypervisor and a basic kernel, with also
includes a driver model that supports all standard drivers for Windows
Server.
Parentpartition
Child partitions
ii ir
■<■■■ Worker
Processes .. User
Applications Applications Applications ,
Windows Server Windows Server Non Hypervisor Xen-enabled
2008 2003 Aware OS Linux Kernel
Windows
Kerne
Kerne
vmbus mode
Windows hypervisor
Ring -
From a naming perspective, what was domain with Xen becomes the
parent partition with Hyper-V. Child partitions correspond to Xen
domains. One big advantage of this structure is that excellent Windows
driver support automatically applies to the virtualization layer. This
makes Hyper-V the server virtualization layer with the broadest driver
support. In addition, Hyper-V uses all the features of the new processor
families from Intel and AMD. There is optimized communication
between Windows Server 2003/Windows Server 2008 and the
hypervisor and physical components. Having a Windows based
hypervisor and using the VMBus make this possible. Even with Linux,
Hyper-V uses special hardware features from Intel (Hypercall Adapter)
to improve the performance of Linux OS based applications running in
Hyper-V virtual machines. This means Hyper-V, in addition to its other
capabilities, fits very nicely into virtualized Linux environments.
The hypervisor itself is available in two options:
Hyper-V Server 2008 R2

(free downloadable)
Microsoft Hyper-VServer
Windows Server 2008 Hyper-V R2

(role of Windows server)
Hyper-V
Contains Windows hypervisor and
other components, including

base kernel and driver technologies
Available as a role in server.

Core or full installation of
Windows Server 2008 R2
Hyper-V has an integrated role in the different versions of Windows

Server 2008 R2. 'Server roles' are server applications that can be
subsequently installed individually. In earlier versions of Microsoft
Windows Server, applications such as Internet Information Services (IIS)
or file release services were configured as standard when Windows
Server was first installed. By contrast, Windows Server 2008 R2 cannot
perform any functions by default. The administrator must first assign
the required roles explicitly to the Server. Virtualization with Hyper-V is
such a role.
From an operating system licensing perspective, the use of Hyper-V as

a role in Windows Server 2008 R2 is often recommended. The reason is
quite simple: One license of Windows Server 2008 R2 Enterprise also
includes the right to run Windows Server in four virtual machines. This
includes the Microsoft licenses for the four guest systems. If Hyper-V
Server is used, an operating system license is required for each virtual
machine. The principal rule from a cost viewpoint is: Use Windows
Server 2008 R2 Datacenter (offered by Fujitsu only in the two processor
version which also can be used for one processor) to run unlimited
virtual machines with the Windows Server operating system on a 2
socket Fujitsu PRIMERGY server.
MicrosoFt
Hyper-VserverR2
World-class virtu alizacirjn of

WS2DD0. HV52Q03, WS2DD8
Linux&Windows clients
Windows Server
2008 R2 Enterprise
Includes virtuafszation rights

for WS 2008 R2
(1 physical + U guests}
Microsoft recognized the importance of broad management around

pure server virtualization technology early on. This was the reason they
finished developing their management product, System Center Virtual
Machine Manager (SCVMM), for virtual machines, even before the
hypervisor was available.
The management of Hyper-V as a role and Hyper-V Server is identical.

The included (or free downloadable) Hyper-V Management Console
and Failover Cluster Manager enable execution of all functions in the
virtualization layer. Moreover, the free downloadable management
suite (Remote Server Administration Tool) can run on a separate PC
(Windows Vista or Windows 7). This enables the same management
functionality to be used remotely.
Page 13 of 32
www.fujitsu.com
For easier and cost-effective Hyper-V virtualization management, use

of SCVMM is recommended. In addition to providing the central library
of the complete virtual environment, an intuitive graphic interface, and
further easy to use add-on's, SCVMM is integrated with Microsoft's
management products in its System Center family.
SCVMM has the following architecture:
Administrator's Self-service
console Web Portal
Windows PowerShell
SDK connection
Operators Web Windows

console console PowerShell
Virtual Machine Manager

server
Operations Manager
server
^W SQLServer r^M
Management interfaces
SAN Storage
Microsoft has included management of VMware environments in its ICT

management architecture. In addition, SCVMM connects to the
Operations Manager; enabling integrated management of both VMs
and real instances (i.e. dedicated real servers and the real instance
behind every virtualized server). The same ICT management also
covers applications, whether they run on a real or a virtual machine. As
a result, Microsoft has created a management solution that can
administer all the resources in a modern datacenter.
Microsoft's mission is not to offer an isolated solution in the virtualized

world, but rather to integrate virtualization management as a further
strategic pillar in its existing comprehensive ICT management portfolio:
the Microsoft System Center family. The most important System Center
products in connection with virtualization are:
■ System Center Virtual Machine Manager (SCVMM)
■ System Center Operations Manager (SCOM)
■ System Center Configuration Manager (SCCM)
■ System Center Data Protection Manager (SCDPM)
■ System Center Essentials (SCE, a special offering for the SME market)
■ System Center Virtual Machine Manager Workgroup Edition (SCVMM

WGE, a special offering for the SME market)
Today the special SME offerings are a combination of SCE (which

consists of key functions from SCOM and SCCM - limited in the number
of supported OS environments) and SCVMM. This is called the System
Center Essentials Management Suite 2007.
This comprehensive solution enables unified proactive and efficient

management of physical and virtual ICT assets. With in-built
troubleshooting knowledge and simplification of the traditionally
complex tasks of update management and software distribution, this
solution helps to significantly reduce the standard administrative tasks.
The management suite also provides functions for easy automation of
standard processes and combines them with the unified management
of virtual machines. SCE2010 will include further functionality
(including VMM functionality) and fewer limitations in the number of
OS environments supported.
Another strength of the Microsoft concept, is the development

expertise at all levels of the ICT infrastructure. This includes the
virtualization layer, the OS, applications, and comprehensive
management of the entire environment. Microsoft can also leverage
this expertise in optimally tuning its own products for this ICT
infrastructure; including the gradual optimization of key applications
(Office and others) for its virtualization layer. Finally, one development
for the entire software infrastructure eliminates compatibility and cross
product support issues that often arises when technology from different
vendors is used.
In addition to standard server virtualization scenarios, Hyper-V is also

well suited to desktop virtualization where XP, Vista or Windows 7 is
used. Here, no explicit Windows Server CALs (Client Access Licenses)
are reguired for the individual virtual machines - each representing a
single desktop user. Instead, RDS (Remote Desktop Services) CALs are
reguired for all desktops accessing the server.
Hyper-V also enables the execution of older versions of Windows Server

OS on new server hardware; even though they are not certified with
older versions of Windows Server due to Microsoft's certification policy.
It is also useful to mention that Microsoft provides a free-of-charge tool

'Disk2vhd' for simple conversion of existing physical Windows Server
environments. This tool enables online migration of a running physical
server onto a virtual machine. However, running this physical server in
a VM does reguire a shutdown of the physical server and a boot of the
virtual machine using the vhd-file produced during 'Disk2vhd'
execution.
Page 14 of 32
www.fujitsu.com
4. Red Hat Enterprise Virtualization (RHEV)
Red Hat started virtualization using the open source product Xen.
However, following the acquisition of XenSource by Citrix, Red Hat
changed its virtualization strategy. In September 2008, Red Hat
announced the acquisition of Qumranet, Inc. The acquisition included
Qumranet's virtualization solutions, its KVM (Kernel Virtual Machine)
platform and SolidlCE, a virtual desktop infrastructure (VDI). Together
they provide a comprehensive virtualization platform for enterprise
customers. In addition, as part of the deal, Qumranet's team of
professionals who were developing, testing and supporting the
solutions, (and leaders in the open source community KVM project)
joined Red Hat. Red Hat subsequently focused on making the KVM
technology, which is actually part of every standard Linux operating
system, into its own virtualization strategy. Their approach uses
kernel-based VMs. This means each VM is mapped in a normal Linux
process that is managed using customary Linux mechanisms. A special
KVM driver is part of the Linux kernel and, among other things,
controls I/O communication between the VMs and the real hardware.
Red Hat Enterprise Virtualization for Servers is an end-to-end

virtualization solution designed for pervasive datacenter virtualization.
The latest Red Hat Enterprise Virtualization introduces several new
features including desktop virtualization with the Red Hat Enterprise
Virtualization for Desktop add-on subscription.
Overview of RHEV Hypervisor:
Red Hat Enterprise Virtualization Hypervisor is based on Kernel-Based

Virtual Machine (KVM) virtualization technology, which can be
deployed either as a standalone bare metal hypervisor (included), or
as part of Red Hat Enterprise Linux 5 and later (purchased separately).
They are installed as a hypervisor host and managed through Red Hat
Enterprise Virtualization Manager for Servers
Secondly, KVM applied a tried and true adage - "don't reinvent the
wheel". Hypervisors need many components in addition to virtualized
CPU and memory. For example, a memory manager, process scheduler,
I/O stack, device drivers, security manager, network stack, etc. In fact,
the hypervisor is really a specialized operating system, the only
difference being it runs virtual machines rather than applications. With
the Linux kernel already including the core features required by a
hypervisor, plus with the maturity and enterprise platform
performance, from over 1 5 years of support and development, it was
more efficient to build on that base rather than write all the same
components again for the hypervisor. In this regard, KVM benefited
from Red Hat's experience with Xen. One of the key challenges of Xen
is the split architecture of domainO and the Xen hypervisor. Since the
Xen hypervisor provides the core platform features within the stack, it
needed to implement the OS-like feature from scratch. However,
because the Linux kernel was already mature and had a proven
memory manager, including support for NUMA (Non-Uniform Memory
Access), large-scale systems, and power management; the decision
was made to incorporate the KVM into the upstream Linux kernel. Red
hat submitted the KVM code to the Linux kernel community in
December of 2006 and acceptance into the 2.6.20 kernel came in
January of 2007. It was at this point that KVM became part of the Linux
core and able to inherit all the key features in the Linux kernel.
KVM Architecture
In KVM architecture, each virtual machine is implemented as a regular

Linux process, and scheduled using the standard Linux scheduler. In
fact, each virtual CPU appears as a regular Linux process. This allows
KVM to benefit from all the features of the Linux kernel. Device
emulation is handled by a modified version of QEMU. This provides
emulated BIOS, PCI bus, USB bus and a standard set of devices,
including: IDE and SCSI disk controllers, network cards, and so on.
KVM is implemented as a loadable kernel module that converts the

Linux kernel into a bare metal hypervisor. There are two key design
principals adopted with KVM. Both have helped it develop rapidly into
a stable and high performance hypervisor, enabling it to overtake
other open source hypervisors.
First, as KVM was designed after the advent of hardware-assisted

virtualization, it did not implement features provided by hardware. The
KVM hypervisor requires Intel VT-X or AMD-V enabled CPUs and
leverages those features in CPU virtualization. By requiring such
hardware support KVM was able to design an optimized hypervisor
solution without the "baggage" of supporting legacy hardware or
requiring modifications to the guest operating system.
Drivers
VM
D" ve~s
VM
VM
Di vers
Hardware
Page 15 of 32
www.fujitsu.com
KVM supports the latest memory virtualization features from CPU

vendors including Intel's Extended Page Table (EPT) and AMD's Rapid
Virtualization Indexing (RVI) for reduced CPU utilization and higher
throughput. Support for memory page sharing comes from a kernel
feature called Kernel Same-page Merging (KSM). KSM scans the
memory of each virtual machine, and VMs that have identical memory
pages are merged into a single page shared between the virtual
machines. If a guest attempts to change this shared page it is given its
own private copy. When many virtual machines are consolidated onto a
host there are many situations where memory pages can be shared -
for example, unused memory within a Windows virtual machine,
common DLLs, libraries, kernels or other objects common between
virtual machines. Using KSM, more virtual machines can be
consolidated on each host, reducing hardware costs and improving
server utilization.
Hardware support
The KVM, as a part of Linux, can leverage the entire hardware

ecosystem. Any hardware device supported by Linux is usable by KVM.
The nature of the open source community means Linux enjoys one of
the largest hardware vendor ecosystems, as new hardware adoption is
rapid. This allows KVM to utilize a wide variety of hardware platforms.
Without additional engineering, KVM inherits each new feature added
to the Linux kernel. Ongoing tuning and optimization of Linux also
immediately benefits KVM.
KVM can also store VM images on any storage supported by Linux,

including IDE, SATAand SCSI local disks, Network Attached Storage
(NAS) including NFS and SAMBA/CIFS or SAN via iSCSI and Fibre
Channel support. Multipath I/O also improves storage throughput and
provides redundancy. Again, with KVM part of the Linux kernel and
support from all leading storage vendors, the KVM storage stack is also
well established in production deployments worldwide.
Virtualization management
The server virtualization management system provides advanced

capabilities for hosts and guests, including high availability, live
migration, storage management, system scheduler, and more. Red Hat
Enterprise Virtualization Manager for Servers is a feature-rich server
virtualization management system that provides advanced capabilities
for hosts and guests, including:
■ Import and Export of Red Hat Enterprise VMs in OVF (open
virtualization format)
■ V2V Conversion Tools to Import Red Hat Enterprise Linux guests from
Red Hat Enterprise Linux/Xen or VMware. Additional converters are
expected will be available in the future
■ Backup and Restore support from leading third-party backup vendors

for Red Hat Enterprise Virtualization VM backup using agents in the
VM
■ Reporting Engine: Documented database schema and sample

reports for extracting information from Red Hat Enterprise
Virtualization Manager
■ Live migration for movement of running virtual machines from one

host to another, with no effect on performance
■ Prioritized high availability where critical VMs automatically restart

on a new host on hardware failure. New priority feature allows the
ranking of VMs to ensure mission critical applications restart as a
priority on the remaining hardware, where number of VMs may be
reduced.
■ Maintenance manager: It allows host upgrades and maintenance

while virtual machines are running
■ Image manager: It enables thin provisioning of disk storage and use

of templates for fast build of new virtual machines from standard
configurations
■ System scheduler: It allows Red Hat Enterprise Virtualization

Manager to migrate virtual machines across multiple hosts to
balance workloads
■ Power saver: It allows Red Hat Enterprise Virtualization Manager to

consolidate virtual machines on a smaller number of hosts during
non-peak hours, plus power down unneeded hosts
■ Search driven user interface with tagging for searching virtual

machines and host characteristics, as well as any assigned arbitrary
tags
■ A consistent management environment for servers and desktops
Page 16 of 32
www.fujitsu.com
5. Oracle VM
Although Oracle entered the market quite recently, Oracle VM already

delivers a large set of functions, even exceeding some of its
competitors. Features like VM high availability, load balancing, live
migration of virtual machines between physical servers and much
more, are available. The technical basis for Oracle VM is identical to the
Xen version of Red Hat; however, Oracle's implementation differs
significantly in the management tools layer, service offering,
integration with Oracle database, applications and middleware
products.
The growing range of offerings in the market raises the question for ICT
organizations of which they should consider. Should it be one for all or
should it be a combination of several depending on use. At a first
glance, it sounds reasonable to go for a one-for-all and
follow-the-market-leader approach.
This is where the use of Oracle VM can be of great help. The complete
stack, above the server hardware, all comes from the one vendor, they
take responsibility for the entire support:
■ OS: Oracle Enterprise Linux
■ The virtualization layer: Oracle VM Server
■ The virtual machine management: Oracle VM Manager, Oracle

Enterprise Manager Grid Control (optional)
■ Compatibility with applications, middleware and databases including

tailored mechanisms for deployment
■ Mix of virtualization technologies (Oracle VM at the server level with

Real Application Cluster at application and database level)
Further advantage comes from the high and increasingly growing

integration with higher layers of the Oracle environment. This is best
illustrated by two examples:
From a pure ICT infrastructure point of view, complexity and operational

costs will certainly be lower by following a one-for-all approach. The
main reasons for using more than one technology mainly result from
price considerations. Today there are still large differences in price and
functionality across the various virtual machine technologies and
management tools. Oracle VM, including Oracle VM Manager, is free of
charge - only service costs are charged. So adopting a one-for-all
approach is typically more expensive from a purchase price point of
view.
Oracle VM templates accelerate the deployment of a fully configured

software stack by offering pre-installed and pre-configured software
images. Proven configurations allow fast roll out and ongoing
maintenance effort is reduced.
Oracle Enterprise Manager proactively monitors the health of all

application components, the hosts they run on, and the key business
processes they support. Together with the Oracle VM Management
Pack, it is now possible to manage, the virtual machines, the OS, and
the software running inside the virtual machines, from a single tool.
Looking at the application environment the picture may change

completely. Experience has shown that not every application behaves
well in a virtual machine environment. Typically, database systems and
transaction-based applications have problems. This can generate
significant integration costs. Another important area to consider is
application vendor support. Support can be quite complex especially if
several vendors are involved for hardware, OS, virtual machine
technology and applications. Quite often, problems are caused by
incompatibilities between different software layers and cannot be
assigned to just one. That is why Oracle provides only limited or no
support if Oracle software is used in virtualization environments other
than Oracle VM.
Finally, there can be commercial issues when it comes to software

licensing. In many ICT organizations, the complete Oracle environment
is quite big, comprising many database systems, several Oracle
applications and numerous deployments of Oracle middleware,
including application servers. These often run on a large number of
servers. Introducing virtualization to consolidate servers and to gain
more flexibility in capacity usage is very appealing. However, all the
potential issues regarding the technical integration of databases and
applications with virtual machines, support issues, license complexity,
and so on, may inhibit starting the virtualization project.
Significant savings in integration effort, from the complete Oracle

stack, well compensate for any rise in complexity from the introduction
of additional virtualization technology. On the operational side, the
higher productivity of the integrated stack provides management from
the virtualization layer up to the application layer; this more than
offsets any additional training costs.
The key strengths delivered by Oracle VM can be summarized as
■ Good price-functionality ratio
■ Good performance from alignment in the development of

virtualization software, middleware and applications
■ End-to-end support within Oracle environments
■ Operational efficiency through integration of management tools
Page 17 of 32
www.fujitsu.com
6. Novell
Novell delivers another implementation of Xen and KVM, providing

users with flexibility and choice in open source virtualization
hypervisors. The Xen and KVM hypervisors and comprehensive
management tools ship as part of SUSE Linux Enterprise Server (SLES)
at no additional cost. Using these integrated capabilities users can
increase resource utilization, provision business services, and improve
business continuity. SUSE Linux Enterprise Virtual Machine Driver Pack
includes a paravirtualized network, plus bus and block device drivers,
to run fully virtualized Windows Server and Red Hat Enterprise Linux
workloads on Xen on top of SLES with near-native performance. Tight
collaboration between Novell and Microsoft has resulted in optimized
cross-platform virtualization -whether running Windows Server guests
on SLES with Xen, or SLES guests on Windows Server with Hyper-V.
Users can also run unlimited SLES guests per physical system. This is for
the price of a single subscription, making it the lowest-cost operating

system for virtual deployment.
As well as integrated Xen and KVM hypervisors, SLES also provides

interoperability with hypervisors from Microsoft and VMware, including
VMware vSphere and Microsoft Hyper-V. SLES includes a VMI-enabled
(VMware's Virtual Machine Interface) kernel, for more efficient
communication with VMware vSphere. This gives better performance
when running as a guest operating system. Microsoft interoperability
also means a more integrated set of Novell and Microsoft management
tools for mixed Linux/Windows and Xen/Hyper-V environments.
Optimized for x86 servers with the latest Intel multi-core architecture,
SLES and the integrated open source hypervisors work well with server
management software. Such combinations can be attractive in the
datacenter.
Server consolidation & virtualization
While many of the servers in today's datacenters are "deemed"

business-critical, most enterprises are yet to embrace virtualization and
migrate their workloads into a virtual environment. This is usually due
to user concerns about performance issues and unplanned downtime.
Novell takes a practical and safe approach to this by first collecting
workload utilization data. Consolidation candidates are then identified,
and consolidation scenarios are generated based on the resources
needed per workload. The workload is migrated over the network onto
a virtual host (P2V) or physical x86 server (P2P) and tested to ensure it
runs as predicted. This has no effect on the existing production server.
When satisfied, any changes that occurred during the test window can
be synchronized and the new virtual workload made live. This
pragmatic, three-step process presents a practical approach to proving
the application will run without impact on business operations. Novell
PlateSpin server consolidation technology is a good fit with large and
very large enterprises looking to migrate hundreds of Windows, Linux
and UNIX, business-critical workloads into a VMware, Xen, or Microsoft
virtual environment.
Datacenter consolidation
Datacenter consolidation involves the planning, migration and

consolidation of hundreds of servers and virtual machines. It can be an
overwhelming task without the proper tools and technology. Major
effort and time is reguired to manually identify, package, and ship
existing servers and virtual hosts to a new site and then set up and
rebuild those same systems. This is therefore a good time to identify
and migrate underutilized servers to a consolidated virtual
environment, plus migrate older servers to new and more powerful
space-saving rack and blade servers.
Virtualization and workload management
Novell PlateSpin workload management is a portfolio of enterprise

class products for physical and virtual server workload handling in
mixed ICT environments. Novell PlateSpin profiles, migrates, protects
and manages workloads over the network between multiple
hypervisors and computing platforms. Designed to fit into existing ICT
infrastructures, it is also able to scale as the datacenter evolves. With
datacenter consolidation and relocation a growing strategy in overall
ICT cost reduction, Novell's virtualization and workload management
enables users to migrate and consolidate Windows, Linux and UNIX
servers, plus VMware, Xen and Microsoft virtual machines - wherever
they are located. Novell PlateSpin works well with enterprises of
different sizes and help reduce costs, complexity and risk. It also helps
in transforming the datacenter into a more agile, efficient and resilient
ICT environment.
Virtualization and workload management from Novell allows reduction

in the manual effort and logistics involved in completing such site
consolidation. This is done by simplifying, automating and planning
the migration of servers and virtual machines over the network to a
more efficient consolidated datacenter location. The solution reduces
risks associated with migrating hundreds of servers and virtual
machines by giving the ability to profile, migrate, and fully test the
workloads in their new environments prior to going live. Such solutions
can migrate hundreds of Windows, Linux and UNIX servers as well as
VMware, Xen and Microsoft virtual machines.
Importantly, once user confidence grows, Novell PlateSpin datacenter

consolidation supports the migration of hundreds of mixed
environment servers and virtual machines across multiple datacenter
locations.
Page 18 of 32
www.fujitsu.com
Summary
VMware today is the benchmark by which all other virtualization

companies are measured. It has the richest feature set, enterprise-class
management tools for virtual infrastructure, and practical tools for
integration into other management and hardware environments. It
also has the highest market share and longest experience. It should
therefore be a top pick where no other virtualization options are
specifically preferred.
The Citrix strategy is an integration ofXenServerwith other products

from their huge portfolio. Such integration offers more advanced
end-to end solutions beyond the hypervisor. In particular, it provides
the longest experience and most complete set of options in desktop
virtualization. If that is also a part of the virtualization eguation then
Citrix needs to be considered.
Microsoft's strength is in the comprehensive management that also

takes care of Microsoft OSs, Microsoft applications, physical servers and
server virtualization technologies other than Hyper-V. If most of the
administrators are Windows trained, then Microsoft virtualization will
make the cleanest and most logical addition to existing Windows
environments.
Red Hat offers an end-to-end solution from KVM based server

virtualization to desktop virtualization. For organizations looking to run
a Linux operation or have a large proportion of Red Hat Linux
environments, this comprehensive set of solutions is a very good fit.
Oracle VM is a Xen based hypervisor with specific focus on serving

Oracle environments. Integrated in Oracle solutions and management
tools, it will deliver the best results to reduce complexity and improve
inter-operability in Oracle environments.
Novell focuses on good interoperability with other vendors (especially

with Microsoft, but also VMware). It has a comprehensive systems
management suite and in combination with PlateSpin technology is
well positioned to provide advanced management in virtual
infrastructures. In mixed environments where no particular
environment dominates, Novell offers a cross platform approach that
also scales well for large server consolidation task.
Page 1 9 of 32 www.fujitsu.com
Hardware support
Optimization for hypervisors through hardware assistants
Depending on the application profile, the use of hypervisor technology

could previously have a remarkably negative impact on performance.
The reasons for this included simultaneous virtual address translation
for multiple virtual machines and I/O emulation.
Today, these problems are being addressed by extensions to the x86

instruction set. In particular, Intel-VT and AMD-V virtualization support
which now all hypervisor products support. Implementation makes
hypervisors simpler, smaller, more efficient and robust. It also allows
users to move more software onto new industry standard server
platforms. Conseguently there is no need to modify the guest OS for
performance reasons, making paravirtualization less important in the
future. However, it needs to be said that hardware assistants in the CPU
chip will not eliminate the need for hypervisors.
x86 operating system architecture
App
App
App
App Ring 3
(Ring 1&2)
Ring
In this architecture, the processor provides 4 privilege levels, also

known as rings. These are arranged in a hierarchical fashion from ring
to ring 3. Ring has the most privileges, full access to the hardware,
and is able to call privileged instructions.
Ring is where the operating system (OS) runs with the OS kernel
controlling access to the underlying hardware. Rings 1, 2 and 3 operate
with lower privilege levels and are prevented from executing
instructions reserved for the operating system. In common operating
systems such as Linux and Windows, the operating system runs in ring
and the user applications run in ring 3. Rings 1 and 2 are not used by
modern commercial operating systems. This architecture ensures that
an application, running in ring 3, cannot make privileged system calls.
However, a an operating system running in ring hardware does
expose applications running in lower privilege levels. This model
provides benefits in traditional, "bare metal" deployments, but it does
present challenges in a virtualized environment.
In a virtualized environment, the hypervisor must run at the most

privileged level, controlling all hardware and system functions.
Ring 3
Hypervisor
(Ring 1&2)
Ring
In this model, the individual virtual machines then run in a lower

privileged ring, typically in ring 3. However, within ring 3 we can see
both virtual machines contain the OS running on virtual (emulated)
hardware. Since the operating system was originally designed to run
directly on the hardware, it expects to be running in ring and will
make privileged calls that are not permitted in ring 3. When these
privileged calls are made, the hardware will trap the instructions and
issue a fault - which typically destroys the virtual machine.
Much of the work performed in an x86 virtualization solution centers
around handling this de-privileging of the operating system running in

the virtual machine. This includes moving the OS kernel from ring to
ring 1 (or higher). This is sometimes referred to as "ring compression".
Binary translation
In this model, pioneered by VMware, instead of emulating the

processor, the virtual machine runs directly on the CPU. Now if privilege
instructions are called, the CPU issues a trap that the hypervisor can
handle by emulation. However, some x86 instructions are not trapped
(e.g. pushf/popf). There are also cases where the virtual machine
identifies it is running in ring 3. To handle these cases, a technigue
called Binary Translation was developed. Here, the hypervisor scans the
virtual machine memory and intercepts such calls before they are
executed and dynamically rewrites the code in memory. The operating
system kernel is unaware of the change and operates normally. This
combination of trap-and-execute and binary translation allows any x86
operating systems to run unmodified on the hypervisor. While this
approach was complex to implement it yielded significant performance
gains compared to fully emulating the CPU.
Page 20 of 32
www.fujitsu.com
Paravirtualization
While emulation and binary-translation focused on how to handle a

privileged instruction executed in a virtual machine, a different
approach was taken first by the open source Xen project. Instead of
handling each privileged instruction as they occurred, the approach
with paravirtualization modified the guest OS in the virtual machine to
replace all privileged instructions with direct calls to the hypervisor. In
this approach, the modified guest operating system is aware it is
running on a hypervisor and can cooperate with the hypervisor for
improved scheduling and I/O. It also removes the need to emulate
hardware devices such as network cards and disk controllers. However,
as paravirtualization reguires changes to the OS, it needs to be
implemented by the operating system vendor. Initially these changes
were made to the Linux operating system, in the form of custom
patches to the Linux kernel. Later they were incorporated into the
released Linux kernel.
To use the performance benefits of paravirtualization outside the Linux

community, the community developed a paravirtualization interface
that supports multiple hypervisors (e.g. Microsoft Hyper-V, VMware
vSphere hypervisor). Using this interface, a paravirtualized Linux
operating system can run on any hypervisor that supports it. As the
same guest OS kernel is able to run natively on the hardware or
virtualized on any compatible hypervisor, transparent
paravirtualization is achieved. This works by design, as all calls have
two implementations, inline native instructions for bare hardware and
indirect calls to a layer between the guest OS and hypervisor for virtual
machines. Applications can be optimized for virtualization in the same
way. Examples are Oracle OVM and Oracle applications or Citrix
XenServer and XenApp.
Using these new hardware features, the hypervisor no longer needs to

implement the binary translation previously reguired to virtualize
privileged instructions. VT-X and AMD-V reduced the overhead when
virtualizing the CPU. However, significant resources are expended by
the hypervisor when handling memory virtualization. Because the
guest operating system cannot directly access memory, the hypervisor
has to implement mapping between the physical host memory and the
virtual memory used by the virtual machine. This is typically
implemented using shadow page tables within the hypervisor.
Recently AMD developed the Rapid Virtualization Indexing (RVI)

feature, previously known as nested page tables. Intel also developed
the Extended Page Table (EPT) feature. Now available with the latest
CPUs, they provide a virtualized memory management unit (MMU) in
hardware. This delivers significant performance improvement
compared to the previous software only implementations. Both Intel
and AMD are continuing to add new features to their processors to
improve virtualization performance, offloading more features from the
hypervisor onto "the silicon" to provide improved performance and a
more robust platform.
The current generation of Intel and AMD CPUs, and chipsets, are now
adding support for I/O offloading. This includes secure PCI
pass-through, using Intel VT-D and AMD I0MMU. This allows PCI devices
on the host to be passed directly into the virtual machine. Single Root
I/O virtualization (SR/IOV) further extends those features to allow
special PCI devices to be split into multiple virtual PCI devices that can
also be passed through to individual virtual machines. These features
are allowing virtual machines to achieve the same I/O performance as
the original bare metal systems.
Hardware assisted virtualization
Eventually, both Intel and AMD developed extensions for their x86
architecture to provide features for use by hypervisor vendors to
simplify CPU virtualization. The first CPUs with these features were
released in late 2005. Today most Intel and AMD CPUs include
hardware virtualization support, including desktop, laptop and server
product lines.
While the implementations by Intel (VT-X) and AMD (AMD-V) are

different, they use a similar approach. A new operating mode is added
to the CPU which now operates in either host or guest mode. The
hypervisor can reguest a process to a CPU that operates in guest mode.
While still seeing the four traditional ring/privilege levels, the CPU is
instructed to trap privileged instructions and then return control to the
hypervisor.
Page 21 of 32
www.fujitsu.com
Rigorous pursuit of this trend, by processor vendors, continues to

deliver even greater agility allowing functions such as VMware vMotion
and Microsoft Live Migration. This is done by:
■ Reducing any loss in performance when hypervisors are used
■ Enhancing the compatibility of individual processor generations
Two examples illustrate these performance improvements:
Reductions in performance loss: The AMD processor "Barcelona" has a

Rapid Virtualization Indexing function that enables optimization of
memory management in virtualized environments. Intel follows the
same direction with its Nehalem architecture and EPT (Extended Page
Tables) technology. In the past the hypervisor managed VM memory
access, but now hardware tables in the processor do it. That
significantly reduces the arithmetic operations in the hypervisor.
VMware vSphere has already implemented all possible virtualization

technology support in the hardware layer and other hypervisors will
follow. Memory-intensive database applications in particular will
benefit from this. Every I/O operation results in activity in the
hypervisor or domain (or parent partition). There the I/O operation is
mapped to the real device. The same happens when the VM receives a
report that an I/O operation is complete. The need for
high-performance I/O handling will lead to greater future cooperation
between I/O card vendors and hypervisor providers, with the aim of
making VMs able to access the real hardware directly. In fact, this could
be seen as a backward step, as the dependency between hardware and
software reappears. However, I/O performance increases significantly.
Ultimately, users will decide where their emphasis lies.
Increased flexibility: vMotion and Live Migration permit a new guality

of service availability. Planned downtime, which accounts for more
than 80% of downtime at a datacenter, can largely be avoided if
running VMs can be moved from one system to another without service
interruption. Up to now, VMs could only be moved within the same
processor family or generation. This restriction forced users to build
server landscapes with homogenous processors - an unacceptable
reguirement given the rapid development of processor technology. This
problem was partially solved by cooperation between hypervisor and
processor vendors. Once again, VMware was first to respond and
developed its enhanced vMotion capability. Together with Intel's VT
Flex Migration and AMD-V Extended Migration technology, is now at
least possible to move VMs between different processor generations
from the same vendor.
Hypervisor embedded in hardware
As implemented in Fujitsu mainframes many years ago, there is now a

clear trend to have virtualization as a default hardware feature. This
means the hypervisor will be integrated into the server hardware, i.e.
locating it in flash memory attached via internal USB. This would allow
certified server hardware to be shipped ready-to-run. From server boot
to running virtual machines would take just a couple of minutes.
Deployment time and administration costs would also be dramatically
reduced. The smaller code footprint would also further reduce the
security exposure. In addition, patching and updating becomes very
simple. With an embedded hypervisor, servers can be kept diskless.
Embedded Hypervisors are also of particular advantage in remote
locations without skilled staff. Moreover, they are the basis for
appliances. Certified Fujitsu PRIMERGY servers can now be supplied
with various embedded hypervisors.
Page 22 of 32
www.fujitsu.com
State of the art
Evolution of virtual machine technology
Virtual machine technology has developed rapidly in recent years.

VMware vSphere provides the best example of how users can improve
flexibility and efficiency of their ICT infrastructures. VMware started with
their hypervisor to address consolidation and isolation. Independence
between hardware and software was also required to run old OSs on
the latest hardware. Later, the focus moved to managing the virtual
infrastructure leading to the VMware release of vCenter.
The objectives of these technology enhancements were as follows:
■ Simple migration of physical to virtual machines
■ Fast provisioning (e.g. by replication of virtual machines and flexible

distribution)
■ Load-balancing (improve server utilization rates and address specific

resource policies with virtual machines)
■ Live migration of virtual machines, including disk allocations from

one physical machine to another
■ Live migration of related file system entries of VMs (more flexible use
of different storage units)
■ Automated high availability for virtual machines (automatic restart

on another server in the event of failure)
■ Enhanced backup of virtual machines
■ Automated patch management for hypervisor and virtual machines

(even for suspended virtual machines)
Today end-to-end solutions are available. Distributed power

management, can deactivate inactive components to reduce energy
consumption costs. This is because most servers consume around 50%
of their peak power even when idle. This also includes many different
types of load balancing. The aim here is all about using the virtualized
VMware vSphere farm and mapping more services onto fewer physical
servers. While at the same time, ensuring that each service has its
resource requirements guaranteed.
VMware Fault Tolerance (FT) is pioneering a new component that

provides continuous application availability, preventing downtime and
data loss in the event of server failure. Built using VMware vLockstep
technology, VMware FT provides operational continuity and high levels
of uptime in VMware vSphere environments. VMware FT works with
existing VMware High Availability (HA) or VMware Distributed Resource
Scheduler (DRS) clusters and can be simply turned on or turned off for
virtual machines. When applications require operational continuity
during critical period-end operations, VMware FT turns on at the click of
a button to provide additional assurance. It is done by creating a
running shadow instance of the primary VM, on another physical
server.
The VMware vLockstep technology keeps both instances in virtual

lockstep. Non-deterministic events on the primary are logged and
transmitted over a gigabit Ethernet network to the shadow. As they
receive identical input, both virtual machines play the exact same set
of events.
Both virtual machines access a common disk and appear as a single

entity, with a single IP address and single MAC address, to other
applications. Only the primary can perform writes. Both virtual
machines constantly check each other's heartbeat and if the secondary
loses the heartbeat of the primary, it immediately takes over. Each
heartbeat is very frequent, with millisecond intervals, allowing
instantaneous failover with no loss of data or status. VMware FT
requires a dedicated network connection, separate from the VMware
vMotion network, between the two physical servers.
VM VM VM
vSphere hypervisor
►j VM 1 1 VM
l ! I
vSphere hypervisor
VM VM VM
vSphere hypervisor
Every vSphere host has its own virtual switch, which manages the
network connection using the build-in LAN cards and present virtual
machines. Virtualization administrators typically manage virtual
machine networking at the physical server level, using virtual switches
inside the hypervisor. Virtual switches require consistent configuring
across many physical hosts, and aligned with the physical network. This
means virtualization and network administrators must work closely
together to ensure proper virtual machine connectivity. Once deployed,
both must continue the close working for joint troubleshooting and
monitoring of the virtualized datacenter.
For new VM's or when moving VM's from one host to another, network
connections need to be adjusted. This means a lot of complex
administration in large environments. VMware vNetwork Distributed
Switch provides a central point of control for such datacenter-level
virtual networking.
Distributed Switch spans many VMware vSphere hosts and aggregates

networking at the datacenter level. VMware vNetwork Distributed
Switch abstracts configuration of individual virtual switches and
enables centralized administration through VMware vCenter Server.
When a new host is added a standard network configuration is applied
to that host. This simplifies the addition of new servers to the existing
environment. Changes to virtual machine networking, across many
physical servers, are also possible centrally. Many users also require
that the security procedures and policies for virtual machines conform
to those of physical servers for security audits.
Page 23 of 32
www.fujitsu.com
VM
NET STATE
VM
NET STATE
VM
NET STATE
VMware vNetwork Distributed Switch
VMwarevSphene
VMwarevSphere
VMwarevSphene
VMware vNetwork Distributed Switch includes support for third-party
virtual switches such as the Cisco Nexus 1000V virtual switch. The Cisco
Nexus 1 000V provides customers with virtual machine granularity and
control using familiar networking tools, allowing end-to-end network
management across both physical and virtual machine environments.
That means that users who have installed Cisco based network
environment have common and uniform network administration.
Another end-to-end solution comes with the development of VMware

vCenter Site Recovery Manager (SRM). The aim here is automatic
restart of the virtualized ICT infrastructure in the backup center, if the
primary datacenter fails. Storage products in conjunction with vCenter
SRM can now deliver a smooth, fully automated process that enables
an orderly restore of the ICT operation with minimum interruption.
The mode of operation and functional range of vCenter SRM are

summarized in the figure below:
Production
Recovery
App
App
App
OS
Hypervisor
App App
App
OS
I Hypervis
w 1
i Simplified and automatic disaster recovery workflows: setup, testing,

failover, and fallback
i Central management of recovery plans from vCenter

i Manual recovery processes become automated recovery plans
i Simplified integration with 3rd party storage replication
i Rapid reliable, manageable, affordable, disaster recovery
It is important to understand that mirroring of the storage system is

reguired for vCenter SRM to work.
DR for desktop virtualization is also being developed (Also refer to the

separate section on desktop virtualization on page 26). This relates to
the following:
■ Avoidance of redundancy in provisioning OS images on expensive

storage units
■ New ways of delivering applications (application streaming)
■ New methods of detecting viruses in centralized desktop

environments.
I/O virtualization
I/O virtualization is about reducing the complexity in local area

networks (LAN) and storage area networks (SAN). Servers
communicate with LAN via network interface controllers (NIC) and with
SAN via host bas adapters (HBA). Both types of controllers are supplied
with static addresses (MAC and WWN respectively). This means that
moving the OS and applications to another server, following a failure,
for example, will reguire reconfiguration of the LAN and/or SAN.
Use of I/O virtualization will overcome these issues. MAC addresses and
WWN as shown to the network must stay fixed.
However, virtualizing the operational I/O addresses will enable dynamic
movement of I/O addresses between servers. If this is fully transparent
to the LAN and SAN domain there is no need to change any network
addresses, or create extra work for the network administrators. This I/O
virtualization reduces management complexity and provides greater
flexibility both when deploying new servers and services and when
replacing failed servers.
Fujitsu ServerView RCVE supports two types of I/O address virtualization:
■ HBA Address rename
■ Fujitsu ServerView Virtual I/O Manager (VIOM) integration
The basic idea of "HBA address rename" is the use of a "logical (virtual)
WWN", which overwrites the fixed (physical) factory-default WWN on
the HBA. The logical (virtual) WWN is used to map the HBA to the
system volume. This method sets the WWN of any spare server's HBA to
the same value as originally set on the production server. This allows
the spare server to connect to and boot from the same boot disk used
by the original production server.
Fujitsu ServerView VIOM takes care of both the virtual MAC (for NIC)
and virtual WWN (for HBA) as part of the server profile assigned to
each production server. If the production server fails, the WWN, MAC
address, boot configuration, and network configuration are simply
inherited by the spare server. Restart is then automatic using the same
SAN disk previously used by the production server.
Page 24 of 32
www.fujitsu.com
The following example illustrates this:
■ Server #1 is supplied with MAC #1 and WWN #1 .
■ Server #2 is supplied with MAC #2 and WWN #2.
■ When server #2 fails, failover to spare server #3 occurs.
■ The virtual connection is "unplugged", and both MAC #2 and WWN #2

addresses are moved to server #3. This requires no changes in any
network.
Normal operation
NIC
HBA
NIC
HBA
MAC #1
WWN #1
I/O address failover
NIC
HBA
MAC #1
WWN #1
NIC
HBA
NIC
HBA
The typical benefits of this approach are:
■ Separation of server management from LAN and SAN management
■ Better provisioning and planning of infrastructure with less cost and

administration effort
■ Setup and pre-configuration of the entire LAN & SAN environment

before any actual server hardware is installed
■ No additional work required to change servers for maintenance or

upgrades
The network and server architecture can be predefined with all

necessary parameters. E.g., virtual WWN and MAC addresses can be
configured in all switches. Normally network administrators have to
wait until a server is delivered to acquire the right WWN or MAC
addresses from their server colleagues. LAN / SAN and server
administrators must then make sure that information is known to the
entire ICT infrastructure. This procedure repeats every time a new server
is delivered. With I/O virtualization, all server parameters for a blade
system, for example, can now be configured at once. Administrators
have less work to do, new servers are available earlier and the risk of
human error reduces. The time required for hardware change is also
decreased which results in better availability and SLAs.
Benefits:
■ Flexible and fast reallocation of application (images) by moving

profiles between server blades and chassis
■ Automated provisioning of server resources based on user

requirements
■ Complete independence from OS and transparent to any application,

network, or storage device
Another user requirement is the need for extra computing power at

specific times - e.g. for reporting services. However, on weekends, the
server normally used for production control could be used for complex
reporting tasks.
The challenges are:
■ Implementing the management instance for the servers and their

connectors
■ Assigning virtual MAC and WWN addresses to different types of LAN /

SAN controllers
■ Server status notification (on, off, or reboot) and re-assignment of

hardware addresses
This is why I/O virtualization was done first with blade servers. The
management blade (Note: Fujitsu PRIMERGY BX400 and BX900 blade
servers have two redundant management blades per chassis) is the
central point for collection, storage and management of each blades
status. Due to the homogenous blade architecture it is easy to
re-assign hardware addresses.
Fujitsu ServerView Virtual I/O Manager provides this I/O virtualization

technology for Fujitsu PRIMERGY BX400 and BX900 blade systems. The
figure below shows the components required:
Fujitsu PRIMERGY BX900 Fujitsu PRIMERGY BX400
ServerView VOIM
I/O connection blades
Page 25 of 32
www.fujitsu.com
Desktop virtualization followed server virtualization
Up to now, virtualization concepts have mostly been applied to server

environments, i.e. server OS and associated applications are the main
virtualization targets. It is however equally possible to virtualize
conventional PC infrastructures in a similar way. This trend is the most
significant PC technology change for years. Virtual Desktop
Infrastructures (VDI) enables each desktop OS to run as a virtual
machine on a server in a datacenter environment. VDI matches three
existing concepts: traditional desktop, traditional server-based
computing, and server virtualization.
VDI gives users the best of all three of these concepts. Virtual desktops
can be individualized and isolated. They also provide considerable
improvements in terms of availability, security, manageability, resource
utilization and energy consumption. One unique benefit of VDI is the
ability gain access to a users' desktop environment from anywhere.
This allows very flexible execution just by logging on at available
terminal. Finally, software lifecycles can be extended, as any previous
dependence between hardware and software is removed.
The addition of complementary technologies enables even greater

efficiency gains. Storage capacity for virtual desktops, as well as the
management of system images and applications can be reduced. The
ability to run applications in isolation from the underlying hardware
removes any incompatibility conflicts. Most importantly, virtual
workplaces can be dynamically created, provisioned and personalized
on demand, with resources only allocated when the user needs them.
At the end of a user session, all resources are released for reuse. This
minimizes infrastructure reguirements and management effort. The
result is a better quality of service level, far greater flexibility and
application performance, in combination with a beneficial reduction in
overall total cost of ownership.
-it 3 -it
--
^
Apps
OS
Apps
OS
Apps
OS
Apps Apps Apps

OS OS OS
Virtualization Layer
Info box
Fujitsu VDI solution
The VDI solution from Fujitsu combines the latest virtualization

technologies with an integrated and complete solution stack of
components. These are carefully tailored to meet each user's
demands. Such an approach frees the organization from evaluating
and testing all technology from scratch, which significantly lowers
the barriers to adoption of new technology. End-to-end services
cover consulting, design, implementation, operation and
maintenance. ROI services for ICT investment decision support help
determine the financial impact of infrastructure changes early in the
project phase. The integrated approach means reduced project time
and less risk for the organization.
Together with its full range of traditional Thin Clients, Fujitsu also
provides a range of innovative Zero Clients. These provide
clutter-free, space-saving productivity as well as complete silence in
the office. While, the low running costs of traditional Thin Clients
result from server based computing to minimize the computing
power (minimal CPU, memory and graphics processing) required at
the desktop; Zero Clients contain no computing power whatsoever.
This revolutionary front-end device has all the benefits of Thin
Clients, but further increases the manageability, security, and cost
benefits, of desktop virtualization.
Fujitsu Zero Client
Traditional
desktop environment
Virtual Desktop
Infrastructure (VDI)
Page 26 of 32
www.fujitsu.com
What is needed?
Long term coexistence of physical and virtual servers will be a reality

in most datacenters
Server virtualization has become a key enabler in building more

dynamic server infrastructures; allowing many users to share the vision
of a flexible and more responsive ICT infrastructure. However, despite

the rapid growth in server virtualization products, the percentage of
x86 server workloads running in virtual machines today is still low.
According to analyst reports, only 20% of all x86 server workloads
currently run on virtual machines. This means, mixed operation of
physical and virtual servers will be a long-term reality in most
datacenters. As a result, the demand for management tools that
manage both physical and virtual environments will continue.
This situation remains very complicated for many users especially with
different hypervisor products in place. Taking advantage of each
hypervisor's specific management tools will mean embracing further
management complexity. The best advice therefore is to introduce a
hypervisor-aware management tool that handles the common basic
daily operations and only drop down to hypervisor specific
management tools for specific functionality.
Fujitsu ServerView Operations Manager
One example is the use of a product like Fujitsu ServerView Operations

Manager. VMware vSphere hypervisor supports the CIM (Common
Information Model) management standard rather than SNMP. Using
the new CIM provider for VMware vSphere hypervisor, Fujitsu
ServerView Operations Manager can manage vSphere hypervisor
-based servers much more reliably and securely than before.
All-important hardware and software data are presented in identical
format to data from Windows or Linux-based systems. However, Fujitsu
ServerView Operations Manager automatically groups the vSphere
hypervisor host and the VMs running on it to simplify administration.
Fujitsu ServerView RAID Manager also supports vSphere hypervisor, and
is integrated in Fujitsu ServerView Operations Manager.
ISetveiV
Status
Q0S
©<
Host Name: hSOQi0-rx3O0s4c.s*fvw*e. 4bgJK.net

HosUP Addre»; 173,17,90.10
[ Option... | | «- Date*! ]
- & System
Fa-,5
^ Temperature Sensors
r^ Power Supply
s^ Power Supply
& Power level
t^roass Storage
{0 ServerView RAID System

<j RAID Adapters
~~ PAID logical oVrves
} System Processors
JBIOSSeNest
Details of RAID Physical Disks:
5tatus Designation S.M.A.R.T.
q? Operational FUJITSU MAX3073RC (0) t*^ Ok
r^ Operational FUJITSU MAX3073RC (4) r^ Ok
nj Operations! FUJITSU MAX5075RC (7) ^J Ok
Fujitsu ServerView Resource Coordinator VE (RCVE)
Dynamic server infrastructure demands wide-ranging know-how to use

the necessary functionality. Administrators need to understand and
manage the mix of physical and virtual servers. They also must master
various virtualization technologies, as well as multi-faceted
combinations of OS and hypervisors. Then of course, there is the

knowledge needed to deal with network addresses, assign server
resources, as well as deploy OS over SAN, and allocate predefined
storage capacity to servers.
Fujitsu ServerView RCVE software was created to overcome the

problems and complexity than can occur in mix infrastructures; as well
as simplify overall management. RCVE standardizes the work reguired
when administering large pools of physical and virtual servers. It helps
reduce the administrative overhead by automating many processes. It
also enables dynamic allocation of server resources while simplifying
high availability operation.
The principle is extremely simple. Administrators define the physical

and virtual resources they wish to assign to applications. Fujitsu
ServerView RCVE then takes care of the rest. Administrators do not need
to acguire in-depth knowledge of the new, often complex technologies
involved - they just use them.
Fujitsu ServerView RCVE consolidates and simplifies administration in

mixed environments and provides uniform control functions that work
the same way with both physical and virtual servers. RCVE supports
blade servers, rack servers, tower servers and even selected servers
from 3rd-party vendors.
With virtualization, the server pool is not a static by dynamic. New

servers can be added as workloads increase and new applications are
rolled out. Servers can be replaced and applications relocated from one
server to another as resource requirements and workloads change. To
enable such rapid allocation of servers, software requires
centralization. This means the OS and applications are no longer
installed on physical servers, but stored as images on a central storage
system. From there the appropriate image can be quickly obtained and
deployed on the relevant server. It makes for very rapid assignment of
new servers.
This is also true with the allocation of LAN and SAN addresses. When
one server assumes the role of another, especially with blade servers,
Fujitsu ServerView RCVE uses the very convenient I/O virtualization
functions of Fujitsu ServerView VIOM. As a result, long system
installation times needed for application relocation become a thing of
the past. Applications that run on multiple servers can be cloned and
then installed on the desired number of servers in a matter of minutes.
Page 27 of 32
www.fujitsu.com
With servers only needing software images assignment, Fujitsu

ServerView RCVE can also simplify server replacement. This is important
for maintenance work and rollout of new software versions. With RCVE
installation, rollout and replacement consumes much less time
compared to conventional environments. Moreover, because both
physical and virtual servers are handled in the same way, mixed
environments operation is far more efficient. Day-to-day work with
physical and virtual servers, even though they use different technology,
can all be managed with the one set of Fujitsu ServerView RCVE tools.
If a server fails, Fujitsu ServerView RCVE automatically starts a spare

system to prevent downtime. This makes it very economical to protect
large numbers of production servers, as the whole pool can be
protected using just one or two substitute servers (N:1 failover). This is
a very cost effective alternative to standard clustering where a
substitute server is reguired for every production server.
Due to the high cost and complexity of clusters only a small portion of
critical applications are typically protected. But such restrictions no long
matter in a N:l environment. Any spare servers can be jointly used in
mixed, physical and virtual environments. In addition, this efficient
high availability function complements existing high availability
functions provided by each hypervisor vendor for their virtual machines
- and can be combined with them if necessary.
r-,.
Fujitsu ServerView Resource Orchestrator (ROR)
The next step in the management of resource pools, is the pooling of

physical servers or a mix of physical and virtual servers in combination
with dynamic provisioning. This is the beginning of cloud computing
where service virtualization is about enabling any service to run on any
server. It can be achieved by storing a complete image of the software
stack (OS, middleware and application) on a central storage system.
Allocation of a server i.e. provisioning of the server, occurs dynamically,
on demand, and just in time. This can happen when the service is
needed, after a failure, or in conjunction with a change in workload. It
means applications can run on servers without any fixed installation
concept. The architecture of this approach is as follows:
Application Application Application
Dynamic Resource Manager
Server
Pool
Storage
Pool
Operating systems
Applications
□ □□□□
Luuuuu
Here the key infrastructure resources, of computing, networking,

storage, and the control of resources, are isolated from each other,
allowing them to be independently scaled and assigned dynamically
The result is a highly adaptable infrastructure that responds to
changing service needs in real time. All resources are available on
demand, providing enterprise-wide agility.
Because there is only one shared OS, installation and deployment is

very rapid. Software updates no longer reguire time-consuming
software distribution. This further reduces complexity and
management effort.
Page 28 of 32
www.fujitsu.com
Fujitsu ServerView ROR enables the provisioning of system resources

with just a few mouse clicks. Whether provisioning a single virtual
machine or a complete server farm with both physical and virtual
servers, everything is very fast. Based on the concept of cloud
computing, users can obtain ICT resources on demand, without
knowing where the resources are, how they are operated, or how the
technology works. Fujitsu ServerView ROR is a good solution for rapid
provisioning and is an enabling technology in building private and
public cloud environments.
Key advantages of Fujitsu ServerView ROR compared to traditional

provisioning are:
■ Quick delivery of system resources to applications
■ Up to 90% time saving in rollout processes
■ Free up of administrative resources
■ Increased use of hardware investments
■ Increased quality of service
■ Improved consistency and system configuration compliance

through standardization and automation
■ Increased agility through fast ICT infrastructure adaptation to

changing business needs
■ SLA-based ICT operation with high levels of service orientation
■ Introduces technology necessary for building cloud environments
Orchestration
Fujitsu ServerView Resource Orchestra tor

Logical Resource Orchestration
Resource
Management
Server Network Storage
Resource Resource Resource
Management Management Management
Abstraction and
Virtualization
Server
Virtualization
I/O Address
Virtualization
Storage
Virtualization
Fujitsu PRIMERGY BX400 and BX900 blade servers
When individual servers are assigned for every application, the typical
result is inefficient use of system resources and high administrative
overheads. The first step in optimization therefore, is the relocation of
applications from isolated servers to pooled architecture. The creation
of pools and the use of server virtualization, significantly increase
resource utilization and management efficiency, and reduce the server
count.
The use of Fujitsu PRIMERGY BX400 and BX900 systems for pool
creation makes handling of the physical servers very simple, especially
as I/O connections and cables are highly consolidated. In addition,
management of both physical and virtual servers is possible with just
one tool - Fujitsu ServerView RCVE. There is also no need to set up
separate resource pools forVMware, Hyper-V, Xen, Linux or Windows.
The one consolidated resource pool rationalizes and leverages all
assets. This also applies to high availability, as the N:1 failover
supported by Fujitsu ServerView RCVE ensures a minimum of standby
systems are required. To summarize:
■ Improved utilization minimizes the number of servers required
■ High availability is achieved at far lower cost
■ ICT administration efficiency increases management productivity
Once the server pool is up and running, all resources become dynamic.
Changes are possible within minutes. Automated operation is also
possible. For example, moving server resources based on schedules
(for day and night or week/weekend operation) or even in response to
sudden changes in workload, becomes straightforward and automatic.
The dynamic server infrastructure created by Fujitsu PRIMERGY blade
servers, Fujitsu ServerView VIOM and Fujitsu ServerView RCVE make
these and other scenarios simple to handle. The main additional
benefits are:
■ The degree of system utilization is further improved
■ Resources are always available for new services at short notice
■ An automatic ability to respond quickly to peak loads
App App App App
Dynamic Resourcing
Virtual server pool
Page 29 of 32
www.fujitsu.com
Fujitsu PRIMERGY BX400 and BX900 fully complement such strategies

and make virtualization and management easy as follows:
■ Physical server blades are pooled by design, with full storage

connection capabilities
■ Complete and versatile LAN & SAN switching capability
■ High performance server blades
■ Memory capacity designed for virtual server pooling
■ Fully certified with all major hypervisors (VMware vSphere, CitrixXen,

Microsoft Hyper-V)
■ High speed mid-plane for hassle-free connectivity and fewer cables
■ Fujitsu ServerView VIOM that takes away administration complexity
■ Complete Scalability, plus Fujitsu ServerView RCVE and ROR, for

dynamic resource orchestration
Fujitsu ETERNUS storage systems
Server virtualization as we have seen is not only a task for server
hardware. Many functions like vMotion, HA, FT, DRS, SRM cannot
happen without a central storage system. A storage system is the
center of any VM farm. The challenge for each storage vendor is to
support virtualization as best as possible. It therefore makes sense to
move storage intensive tasks of the hypervisor and perform them
directly in the storage system. To enable this, hypervisor vendors are
delivering APIs to allow greater interaction with storage system.
For example, VMware delivers vStorage APIs that have different

functions and license editions for its vSphere hypervisor.
■ vStorage APIs for Data Protection
■ Storage vMotion
■ vStorage APIs for Multipathing
■ vStorage APIs for Array Integration
vStorage APIs for Data Protection are for use by enterprise backup
solutions to secure virtual machines during runtime. They allow
administrators to execute scalable backup without disrupting
applications or users. Fujitsu ETERNUS storage systems with vStorage
APIs for Data Protection become a very effective way to enable general
backup in medium to large datacenter environments.
Storage vMotion enables live migration of virtual disks from one

storage location to another without disruption to users. Storage
vMotion also minimizes the need to schedule application downtime
reguired for storage maintenance. VMware Storage vMotion enables
live migration of virtual machine disk files across storage arrays or
operating LUNs. Now it is possible to change/extend storage systems,
optimize memory I/O performance, service, or migration to another
storage system, during operation.
In addition the high performance Fujitsu ETERNUS DX400/DX8000 disk

storage systems support the following three functions:
■ Full copy: In standard virtual machine deployment or cloning

operations, data needs to be copied using vSphere servers. Now the
storage systems are able to make full copies of data within the
storage system environment without a vSphere server to read and
write data. The result is both reduced server workloads and improved
resource utilization effectiveness.
■ Block zeroing: When allocating storage space to a new virtual

machine, the initialization reguires the writing of zeroes to fill the
space. This zeroing process can now be offloaded to the disk storage
systems, eliminating the need for vSphere servers to zero out large
numbers of storage system blocks. This reduces server workloads and
speeds up virtual machine provisioning.
■ Hardware assisted locking: When exclusive data access is reguired,

locking is now possible at the block level, as opposed to the
traditional LUN levels. The result is improved VM operational
effectiveness by minimization of the locking range reguired for
access restrictions.
Page 30 of 32
www.fujitsu.com
Conclusion
Virtualization is a game changing technology. In a world where it is

now clear that every piece of information and every process will
eventually be digitized, virtualization provides an opportunity to take
stock and prepare for a much more computing intensive future.
Importantly virtualization can deliver many short-term benefits whilst
organization prepare for that future. This includes cost reductions from
using virtualization to improve the use of the computing power that is
already available.
Organizations will also greatly benefit from consolidating their older

distributed computing environments onto fewer more powerful and
more efficiently used machines. The centralization that this enables
also has the potential to allow reductions in administration costs as
well as improved security. Moving to a virtualized environment allows
the organization to take advantage of cloud computing. That is the
next logical step in virtualization and has the potential to deliver even
greater economies of scale in ICT provisioning and management.
However, as with all change, the benefits need to be carefully

evaluated against any possible downside. Hypervisors reduce
complexity in physical environments, but also create a new complexity
at the virtual level. Virtualization brings the same risk created by open
systems. Open systems flexibility enabled an explosion in server
deployments. The same could occur with virtual server instances,
especially if strict policies are not put in place to avoid virtual server
sprawl.
Added value in server virtualization, from Fujitsu
Fujitsu is using its experience to embrace virtualization in ways that go

beyond just delivering technology. It bases its approach on working
with customers as partners, and ensuring that virtualization adoption
integrates into the overall use and functions of the organization. It also
focuses on achieving the best possible administration of the resulting
ICT resources.
Fujitsu's Dynamic Infrastructures strategy, which embraces all aspects

of ICT infrastructure including virtualization, follows this approach. This
includes the provision of functions that take care of the new
complexities of managing both physical and virtual environments in
seamless and automated ways.
Fujitsu is also active in the development of cloud computing and is

ideally placed to provide advice on virtualization projects with a view to
cloud computing adopting in the future. Fujitsu delivers the hardware
platforms, software from its own development as well as partner
software, and integrates them all into complete infrastructure solutions
that maximize the benefits of a virtualization approach.
The overall object is to release each organization from the need to

spend too much of their own resources on designing, building, running
and maintaining the ICT infrastructure. This in turn provides the ability
to spend more on supporting business processes through application
development and improved implementation.
Each organization will need to evaluate the following:
■ Management of both virtual and real resources despite the real

differences in management requirements
■ Applications that require special software stacks: Using different

hypervisors can lead to compatibility and support problems.
Examples are applications from SAP, Oracle and Microsoft. Oracle
software is not certified to run on VMware for example. Analysis of
hypervisor collaboration with OSs and applications, leads to complex
combinations.
■ The virtual machine concept is no use if an application already uses

all the server resources in a native, real environment. Examples are
applications with high I/O needs, such as DB servers or mail servers.
■ There will be a need for additional layers of virtualization beyond the

server to be fully integrated.
A major success factor will be achieving a uniform and simple

management of both the real and virtual resources, with the ability to
handle the complexity of virtualization easily, transparently and at an
affordable cost.
Page 31 of 32
www.fujitsu.com
Contact © Copyright 201 1 Fujitsu limited. Fujitsu, the Fujitsu logo are trademarks or registered tra
demarks of
FUJITSU LIMITED Fujitsu Limited in Japan and other countries. Other company, product and service name
s may be
Website: www.fujitsu.com trademarks or registered trademarks of their respective owners. Technical da

ta subject to modification
201 1-03-01 WW-EN and delivery subject to availability. Any liability that the data and illustrations
are complete, actual or
correct is excluded. Designations may be trademarks and/or copyrights of the respective manufacturer,
the use of which by third parties for their own purposes may infringe the rights of such owner.
Page 32 of 32 www.fujitsu.com

Full Text of - Virtualization at Fujitsu

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Full Text of - Virtualization at Fujitsu

Uploaded by

Copyright:

Available Formats

5/9/2019 Full text of "Virtualization at Fujitsu"

Full text of "Virtualization at Fujitsu"

White paper Virtualization at Fujitsu

White paper Virtualization at Fujitsu

Broadly speaking, virtualization in an information and communication

Virtualization is not new

Virtualization, introduced more than 30 years ago with mainframe

■ With server hardware extremely expensive at the time, it needed to

■ By separation of the application reguirements from the restrictions of

The technology developed at that time corresponds very closely to the

Server virtualization is rooted in mature mainframe technology

Fujitsu was one of the virtualization pioneers, and has decades of

Today the approach to virtualization and the levels at which it is

1 Microsoft SoftGrid, Altiris Software Visualization Solution,

Hmj^HmH VMware View, Citrix XenDesktopServer, ...

1 VMware {GSX) Server, Microsoft Virtual Server, KVM...

1 VMware ESX Server, Xen, Microsoft Hyper-V

1 Fujitsu ETERNUSCS, Fujitsu ETERNUSDX, EMC, NetApp,

The wide range of virtualization solutions however means that a key

This white paper discusses the best-known technologies, shows their

White paper Virtualization at Fujitsu

Server virtualization with x86 servers

No other server architecture has gained in importance over the past

The traditional x86 platform approach

The architecture of a traditional industry standard server consists of

■ First is the hardware layer, which is a set of different devices and

■ Second is the OS layer, which has overall control of hardware

■ Third is the applications/services layer, which uses the OS as an

In this traditional platform approach, each configuration layer is tightly

Due to historical reasons, OS resource management is not very efficient

sever utilization. Poor server utilization, in its turn, leads on to an

The overall result is management complexity and low efficiency, with

x86 platform virtualization

To overcome the traditional challenges created by the strong

By definition, server virtualization means hiding the server resources,

Hypervisor technology lets multiple instances of an OS and associated

The traditional platform approach therefore becomes synonymous with

White paper Virtualization at Fujitsu

Virtual machines in OS hosted environments

Some virtualization technology is used less with today's x86 servers. OS

The virtual machine concept in OS hosted environments came into

Virtualization Layer (VMM]

Host OS (Windows or Linux)

■ Protection: Number five with large enterprises, it rises to number two

■ Deployment: Most mid-sized businesses have limited administrative

■ Freedom of choice: Virtualization allows purchases based on

■ Operational agility: Here virtualization helps in adapting server

Overall, mid-sized businesses, while also driven by the promise of

Customer expectations of the virtual machine concept

When it comes to the question of what customers expect from the

■ Consolidation: Which provides the opportunity for dramatic capital

■ Virtualization: As an enabler in reducing the administrative burden of

■ Operational agility: The ability to move workloads from one server to

■ Freedom of choice: Virtualization makes it easier to accommodate a

■ Protection: Virtualization simplifies disaster recovery - but only with

However due to the heterogeneous nature of most large enterprise ICT

White paper Virtualization at Fujitsu

By definition, a virtual machine is a software-driven implementation of

overall underlying equipment in the physical machine.

Four key features characterize the virtual machine concept.

Second, the overall physical server hardware is partitioned into several

Third, with the hypervisor situated between the software stack

Major university in Tokyo